AIX becomes unresponsive on "amf_get_psinfo" when running VCS cluster

book

Article ID: 100014010

calendar_today

Updated On:

Description

Workaround

Disable Process offline monitoring with AMF. For detailed steps, refer to https://www.veritas.com/docs/000022623.

 

 

    Cause

    In AIX 7.1 TL3 SP3 (and AIX 6.1TL9 SP3),  IBM has made a fix that prohibits reading (blocks read calls) process info from the /proc//psinfo file, when a process is in EXEC state.
    In AMF, we get a callback from AIX whenever a new process is placed into the EXEC state. In this context, we read the process information to match with the registered resource information for the VCS agent. This causes a deadlock, as a read call is waiting for EXEC to complete and the EXEC context is waiting on read.
     

    Refer to AIX APAR IV52781 for detailed information:
    IV52781: IF A PROC STOPPED BY PCSEXIT /PROC//PSINFO RETURNS ERROR. APPLIES TO AIX 7100-03
    http://www-01.ibm.com/support/docview.wss?uid=isg1IV52781

    Note: AMF does not function on earlier AIX releases, as well if the user applies this specific APAR.

    Resolution

    Install the IBM APAR IV63274 on all nodes in the cluster to fix this issue:
    https://www.ibm.com/support/docview.wss?uid=isg1IV63274

     

    Issue/Introduction

    AIX becomes unresponsive after starting VCS (Veritas Cluster Server) and bringing a service group online.
    Applies to:
    • AIX 7.1 TL3 SP3
    • AIX 6.1 TL9 SP3
    The following SFHA releases will be impacted if AMF is enabled on these versions of AIX:
    • 5.1 SP1
    • 6.0
    • 6.1
    • 6.2